Codex would

ryandrake 4 hours ago | parent | next [-]

It's crazy. Reverse engineering a protocol/interface is something LLMs are surprisingly good at, and these show a promise of unlocking sooo much proprietary crapware that's hidden behind obfuscation, accounts, subscriptions, lazy vendors who only write crappy windows programs, and so on.

But of course, the major US labs are going to get pressure from these vendors whose business models depend on obfuscation, and you know they're going to bend the knee and nerf the LLMs.

▲

qingcharles 6 hours ago | parent | prev [-]

Not these days. It won't even look at my projects for security flaws any longer. Has a little pop-up that says "This output cannot be viewed." or some such.

Have to make the Chinese one do it.

▲

KlutzySofa 5 hours ago | parent | next [-]

Its backwards in so many ways.

You fall back on having to use Chinese models. In theory the Chinese models could also be trained to sneak in new security flaws while Claude and Codex would avoid looking for them.

▲

theplumber 5 hours ago | parent [-]

Claude and codex could also be trained to sneak security flaws. They could even target specific people. Not to mention that all their data goes through NSA headquarters so you don’t even have the option to run it locally in any shape or form.

	▲	KlutzySofa 4 hours ago \| parent [-]
		The US companies are doing this to not get export restricted by the US government which in turn can end up with companies using Chinese models to get around those restrictions. Its not really a matter of who's good and bad (neither the US or China has the best track record), but creating a scenario where the other party can use your own policies against you.

▲

mrinf1nity 4 hours ago | parent | prev | next [-]

My lil' Claude tells me that my security related request is flagged and will be sent for human review. And to use a lesser model.

I told it 'Let's find the limitations of this Linux system' then it made a benchmarking set of tools for disk i/o, RAM, etc

When using Gemini on my phone it also told me it'll be 'human reviewed.' when asking it about car prices.

Also when saying 'let's use best security practices'

o_O

▲

drnick1 4 hours ago | parent | prev | next [-]

Just tell it to continue. For now, it still mostly works. On the other hand, Claude flat out refuses to look at game and anti-cheat binaries, etc.

	▲	qingcharles 3 hours ago \| parent [-]
		It sorta, sometimes does. It's not a reliable fix, though, sadly. Even my old reliable Muse Spark has gotten a little testy lately. I'm back to GLM 5.3.

▲

catlifeonmars 4 hours ago | parent | prev | next [-]

Get the Chinese one to jailbreak the proprietary one. Best of both worlds.

▲

nullsanity 5 hours ago | parent | prev [-]

[dead]